Are Ansible handlers defined in roles ran after the entire playbook or the role?

Solution 1:

Handlers are executed:

  • at the end of a play (not playbook)
  • on executing the meta: flush_handlers task

So "to add a 6 role to the end that needs to have the handlers of the 4th role" you need:

  • either to split the role assignment into separate plays;
  • or add a meta task and include the 6th role with include_role module:

    roles:
      - role4
    tasks:
      - meta: flush_handlers
      - include_role:
          name: role6
    

For your use case, I'd suggest the first method as the include_role module is still very fresh and there are quirks when using it (see this question on SO).


Moreover, please notice that handlers' names and listen calls are global, so two handlers in separate roles will be in conflict if they had the same name and both roles were assigned in a single play. (ref. Handlers: Running Operations On Change)

Handlers [ ] are referenced by a globally unique name, and are notified by notifiers. [ ] a handler, it will run only once, after all of the tasks complete in a particular play.

Handler names and listen topics live in a global namespace.


  • Empirical proof (run this shell script to confirm handlers are executed at the end of the play - there were contradicting comments and answers here):

    #!/bin/bash
    
    mkdir -p ./sf831880/roles/role1
    mkdir -p ./sf831880/roles/role1/handlers
    mkdir -p ./sf831880/roles/role1/tasks
    mkdir -p ./sf831880/roles/role2
    mkdir -p ./sf831880/roles/role2/handlers
    mkdir -p ./sf831880/roles/role2/tasks
    
    cat >./sf831880/roles/role1/tasks/main.yml <<TASKS1_END
    ---
    - name: Always true in role1
      command: echo role1
      notify: handler1
    TASKS1_END
    
    cat >./sf831880/roles/role2/tasks/main.yml <<TASKS2_END
    ---
    - name: Always true in role2
      command: echo role2
      notify: handler2
    TASKS2_END
    
    cat >./sf831880/roles/role1/handlers/main.yml <<HANDLERS1_END
    ---
    - name: handler1
      debug:
        msg: "This is a handler in role1"
    HANDLERS1_END
    
    cat >./sf831880/roles/role2/handlers/main.yml <<HANDLERS2_END
    ---
    - name: handler2
      debug:
        msg: "This is a handler in role2"
    HANDLERS2_END
    
    cat >./sf831880/playbook.yml <<PLAYBOOK_END
    ---
    - hosts: localhost
      gather_facts: no
      connection: local
      roles:
        - role1
        - role2
      tasks:
        - debug:
            msg: "This is a task in a play"
    PLAYBOOK_END
    
    ansible-playbook ./sf831880/playbook.yml
    

    Result:

    PLAY [localhost] ***************************************************************
    
    TASK [role1 : Always true in role1] ********************************************
    changed: [localhost]
    
    TASK [role2 : Always true in role2] ********************************************
    changed: [localhost]
    
    TASK [debug] *******************************************************************
    ok: [localhost] => {
        "msg": "This is a task in a play"
    }
    
    RUNNING HANDLER [role1 : handler1] *********************************************
    ok: [localhost] => {
        "msg": "This is a handler in role1"
    }
    
    RUNNING HANDLER [role2 : handler2] *********************************************
    ok: [localhost] => {
        "msg": "This is a handler in role2"
    
  • Play modified to contain meta: flush_handlers:

    ---
    - hosts: localhost
      gather_facts: no
      connection: local
      roles:
        - role1
        - role2
      tasks:
        - meta: flush_handlers
        - debug:
            msg: "This is a task in a play"
    

    The result:

    PLAY [localhost] ***************************************************************
    
    TASK [role1 : Always true in role1] ********************************************
    changed: [localhost]
    
    TASK [role2 : Always true in role2] ********************************************
    changed: [localhost]
    
    RUNNING HANDLER [role1 : handler1] *********************************************
    ok: [localhost] => {
        "msg": "This is a handler in role1"
    }
    
    RUNNING HANDLER [role2 : handler2] *********************************************
    ok: [localhost] => {
        "msg": "This is a handler in role2"
    }
    
    TASK [debug] *******************************************************************
    ok: [localhost] => {
        "msg": "This is a task in a play"
    

Solution 2:

Handlers are lists of tasks, not really any different from regular tasks, that are referenced by a globally unique name, and are notified by notifiers. If nothing notifies a handler, it will not run. Regardless of how many tasks notify a handler, it will run only once, after all of the tasks complete in a particular play. ansible doc

1) Handlers that do the same thing should be named the same.
restart nginx ALWAYS restarts nginx, not handler1 and handler2

2) Handlers are run at the END of the entire "Play" a play scoped to your sections.

3) I would use the register and when functions for tasks that should be restarted, note this var should carry with you.

Code Source

PLAY [localhost] ***************************************************************

TASK [debug] *******************************************************************
ok: [localhost] => {
    "msg": "Play 1"
}

TASK [role1 : Always true in role1] ********************************************
changed: [localhost]

TASK [role1 : Always true in role1] ********************************************
changed: [localhost]

TASK [role1 : Always true in role1] ********************************************
changed: [localhost]

TASK [role1 : Always true in role1] ********************************************
changed: [localhost]

TASK [role1 : Always true in role1] ********************************************
changed: [localhost]

TASK [role2 : Run if change in task c of role 1] *******************************
changed: [localhost]

TASK [role2 : Always true in role2] ********************************************
changed: [localhost]

TASK [debug] *******************************************************************
ok: [localhost] => {
    "msg": "This is a task in a play"
}

RUNNING HANDLER [role1 : handler] **********************************************
ok: [localhost] => {
    "msg": "This is a handler in role1"
}

PLAY [localhost] ***************************************************************

TASK [debug] *******************************************************************
ok: [localhost] => {
    "msg": "Play 2"
}

TASK [role1 : Always true in role1] ********************************************
changed: [localhost]

TASK [role1 : Always true in role1] ********************************************
changed: [localhost]

TASK [role1 : Always true in role1] ********************************************
changed: [localhost]

TASK [role1 : Always true in role1] ********************************************
changed: [localhost]

TASK [role1 : Always true in role1] ********************************************
changed: [localhost]

TASK [role2 : Run if change in task c of role 1] *******************************
changed: [localhost]

TASK [role2 : Always true in role2] ********************************************
changed: [localhost]

TASK [debug] *******************************************************************
ok: [localhost] => {
    "msg": "This is a task in a play"
}

RUNNING HANDLER [role1 : handler] **********************************************
ok: [localhost] => {
    "msg": "This is a handler in role1"
}

PLAY RECAP *********************************************************************
localhost                  : ok=20   changed=14   unreachable=0    failed=0

Lots of ways to do the same task. Handlers were designed to prevent restarting the same process multiple times, such as multiple changes to a nginx server that has websites, ssl certs, and other tasks that need service restarts.

Tags:

Ansible